For Want of a Modem and a Comfortable Chair: A Research Note

Abstract

A significant body of work concludes that money matters little, if at all, when it comes to educational achievement (Hanushek 1989, 1994a, 1994b). Others have challenged that view (Hedges, Laine, and Greenwald 1994), and have sought to demonstrate that in some circumstances, and for some things, money does matter. That debate seems to have settled around the idea that if spending money is to make a difference we need to identify those areas in which it might be spent most effectively. Despite an emphasis on examining potential sources of performance variation, very little systematic work has been done with regard to the possible impact of infrastructure on educational achievement. Utilizing statelevel data from a June 1996 GAO report which profiles the condition of American public school facilities by state, this study sought to answer the question of whether infrastructure matters, whether between state variations in the condition of school facilities are related to interstate variations in performance. A factor analysis using oblique rotation identifies three factors which summarize facilities variables in the GAO report: Factor 1-Insufficient technological capabilities, Factor 2-Unsatisfactory environmental conditions, and Factor 3-Inadequate building structures. These three factors are then examined in a series of OLS and SWLS regression (Meier and Keiser 1996) equations designed to test hypotheses about the relationship between public school facilities and systemwide performance. In all equations the dependent variable (SE195) is a standardized measure of each state's educational performance. SE195 is calculated by taking the state mean SAT or ACT score-only if the test was taken by at least 29% of eligible test takers I-and expressing it as a percentage of the highest possible score on each exam. Where dual state scores exist, the average of SAT and ACT means represents that state's score. All models in the study also contain a single control variable-SE188 (calculated in the same manner as SE195). This control is essentially a seven-year lagged version of the dependent variable and accomplishes two